* [lm-sensors] RTC + DS1337
2006-06-29 7:06 [lm-sensors] RTC + DS1337 Matthieu TORRES
@ 2006-06-29 9:43 ` Michael Burian
2006-06-29 10:06 ` Rudolf Marek
` (2 subsequent siblings)
3 siblings, 0 replies; 5+ messages in thread
From: Michael Burian @ 2006-06-29 9:43 UTC (permalink / raw)
To: lm-sensors
Matthieu TORRES wrote:
>
> - the option "Enhanced Real Time Clock Support" (config RTC) does not
> provide generic access to a RTC chip via /dev/rtc as I belevied. The
> access to the chip in driver/char/rtc.c are made with CMOS_READ() and
> CMOS_WRITE(), defined in Linux/include/asm-arm/mc146818rtc.h. It is a
> combination of outb_p() and inb_p() : no chance to access my I2C chip
> that way...
> So I added ioct() support in /drivers/i2c/chips/ds1337.c.
>
I remember having fixed that the same way some time ago - but people
told me that I shouldn't add that code to ds1337.c and to look for a
"generic way" instead.
As I haven't found a better way yet, I decided to write some glue code
(drivers/char/ds1337glue.c) till the issue can be resolved.
http://www.kernel.org/git/?p=linux/kernel/git/mburian/linux-2.6-ep93xx.git;a=blob;hÇd168b7ea488676106a363311dc6119e61519a2;hbá6e2f3f0a126eccf47d4d338b3b1a9281b7cd87;f=drivers/char/ds1337glue.c
I doubt that this will be acceptable for mainline, so I'm still waiting
for somebody (Jean?) to show me the "one true way" to get /dev/rtc
access for that chip.
> - the driver (/drivers/i2c/chips/ds1337.c) has two problems (I'm not
> sure it is bugs, but at least on my system it does not work).
> First, ds1337_detect() reads some values in the chip registers that are
> assumed to be wrong, but in my chip they do exist (I swear I have a
> DS1339). I commented out these checks.
> Then, the part that enables the RTC after first power up in
> ds1337_init_client() fails to write a status bit in the chip, so we fall
> into the same reset code (that writes zeros in the RTC regs) after each
> reboot.
> I corrected this and that's OK for me.
>
> I can post code if it is usefull for someone.
I've already sent a patch that should fix this in February :(
http://lists.lm-sensors.org/pipermail/lm-sensors/2006-February/015369.html
I'm sending a "re-diffed against 2.6.17" version of that patch that,
please review and test.
To Matthieu: If your fix looks completely different, is better, ... or
you'd just like to discuss it, please show us your version.
I can live with whatever i2c folks here prefer, as long as it fixes the
problem.
This patch no longer overwrites the the complete RTC in case it detects
inconsistencies - instead it does just what's absolutely necessary to
enable userspace to fix the problem.
--- linux-2.6.17.old/drivers/i2c/chips/ds1337.c 2006-06-18
03:49:35.000000000 +0200
+++ linux-2.6.17/drivers/i2c/chips/ds1337.c 2006-06-29
10:49:48.000000000 +0200
@@ -337,38 +337,36 @@ exit:
static void ds1337_init_client(struct i2c_client *client)
{
+ s32 val;
u8 status, control;
/* On some boards, the RTC isn't configured by boot firmware.
* Handle that case by starting/configuring the RTC now.
*/
- status = i2c_smbus_read_byte_data(client, DS1337_REG_STATUS);
control = i2c_smbus_read_byte_data(client, DS1337_REG_CONTROL);
-
- if ((status & 0x80) || (control & 0x80)) {
+ if (control & 0x80) {
/* RTC not running */
- u8 buf[16];
- struct i2c_msg msg[1];
-
dev_dbg(&client->dev, "%s: RTC not running!\n",
__FUNCTION__);
+ /* reset EOSC to start RTC */
+ i2c_smbus_write_byte_data(client, DS1337_REG_CONTROL,
+ control & 0x7f);
+ }
- /* Initialize all, including STATUS and CONTROL to zero */
- memset(buf, 0, sizeof(buf));
- msg[0].addr = client->addr;
- msg[0].flags = 0;
- msg[0].len = sizeof(buf);
- msg[0].buf = &buf[0];
-
- i2c_transfer(client->adapter, msg, 1);
- } else {
- /* Running: ensure that device is set in 24-hour mode */
- s32 val;
-
- val = i2c_smbus_read_byte_data(client, DS1337_REG_HOUR);
- if ((val >= 0) && (val & (1 << 6)))
- i2c_smbus_write_byte_data(client, DS1337_REG_HOUR,
- val & 0x3f);
+ /* Check if the oscillator was stopped due to power failure, ... */
+ status = i2c_smbus_read_byte_data(client, DS1337_REG_STATUS);
+ if (status & 0x80) {
+ dev_dbg(&client->dev, "%s: oscillator was stopped, RTC data"
+ " unreliable!\n", __FUNCTION__);
+ /* reset OSF */
+ i2c_smbus_write_byte_data(client, DS1337_REG_STATUS,
+ status & 0x7f);
}
+
+ /* Ensure that device is set in 24-hour mode */
+ val = i2c_smbus_read_byte_data(client, DS1337_REG_HOUR);
+ if ((val >= 0) && (val & (1 << 6)))
+ i2c_smbus_write_byte_data(client, DS1337_REG_HOUR,
+ val & 0x3f);
}
static int ds1337_detach_client(struct i2c_client *client)
^ permalink raw reply [flat|nested] 5+ messages in thread* [lm-sensors] RTC + DS1337
2006-06-29 7:06 [lm-sensors] RTC + DS1337 Matthieu TORRES
2006-06-29 9:43 ` Michael Burian
2006-06-29 10:06 ` Rudolf Marek
@ 2006-06-29 11:34 ` Ben Dooks
2006-06-30 8:03 ` [lm-sensors] [i2c] " James Chapman
3 siblings, 0 replies; 5+ messages in thread
From: Ben Dooks @ 2006-06-29 11:34 UTC (permalink / raw)
To: lm-sensors
On Thu, Jun 29, 2006 at 09:06:54AM +0200, Matthieu TORRES wrote:
> Hi,
>
> I faced a problem with DS1337 RTC. I use a pxa255 cpu with 2.6.16
> kernel. I tried to make a DS1339 chip (real time clock) working with
> hwclock.
> In my config, I selected I2C, DS1337/DS1339 driver, and ehanced RTC
> support to have it working with /dev/rtc as expected by hwclock. That is
> to say the following options:
>
> CONFIG_RTC=y
> CONFIG_I2C=y
> CONFIG_I2C_PXA=y
> CONFIG_SENSORS_DS1337=y
>
> The ioctl() to dev/rtc did not work. While investaigating on that issue,
> I found 2 problems :
Look at 2.6.17, which has a new RTC core, and sysfs based interface
which is much cleaner way of dealing with the RTC systems on the many
different systems that there are.
--
Ben (ben at fluff.org, http://www.fluff.org/)
'a smiley only costs 4 bytes'
^ permalink raw reply [flat|nested] 5+ messages in thread
* [lm-sensors] [i2c] RTC + DS1337
2006-06-29 7:06 [lm-sensors] RTC + DS1337 Matthieu TORRES
` (2 preceding siblings ...)
2006-06-29 11:34 ` Ben Dooks
@ 2006-06-30 8:03 ` James Chapman
3 siblings, 0 replies; 5+ messages in thread
From: James Chapman @ 2006-06-30 8:03 UTC (permalink / raw)
To: lm-sensors
Hi, there's a new driver in the -mm tree for use with the new RTC core.
It's based on the original ds1337/9 driver and now also supports ds1307.
It was added in the 2.6.17-rc1 timeframe. Look for drivers/rtc/rtc-ds1307.c.
Ben Dooks wrote:
> On Thu, Jun 29, 2006 at 09:06:54AM +0200, Matthieu TORRES wrote:
>
>>Hi,
>>
>>I faced a problem with DS1337 RTC. I use a pxa255 cpu with 2.6.16
>>kernel. I tried to make a DS1339 chip (real time clock) working with
>>hwclock.
>>In my config, I selected I2C, DS1337/DS1339 driver, and ehanced RTC
>>support to have it working with /dev/rtc as expected by hwclock. That is
>>to say the following options:
>>
>>CONFIG_RTC=y
>>CONFIG_I2C=y
>>CONFIG_I2C_PXA=y
>>CONFIG_SENSORS_DS1337=y
>>
>>The ioctl() to dev/rtc did not work. While investaigating on that issue,
>>I found 2 problems :
>
>
> Look at 2.6.17, which has a new RTC core, and sysfs based interface
> which is much cleaner way of dealing with the RTC systems on the many
> different systems that there are.
>
^ permalink raw reply [flat|nested] 5+ messages in thread